On Thu, Aug 07, 2014 at 08:21:14PM +0200, Pascal Hambourg wrote: > > Kernel modules that are needed for the boot process itself reside in the > > initrd, and he copies that. > > For the *early* boot process. I.e. mounting the root filesystem and not > much else.
Indeed. So, considering worst case scenario one will have mounted root filesystem and a console access to the host. That's much better than trying to do anything with the system from initrd's busybox IMO. > > > Kernel modules that live in /lib/modules are loaded after root > > filesystem is mounted and init is started. As long as kernel's ABI isn't > > changed they should load successfully. > > But this method provides no backup if the update causes a regression in > one of these modules which may break a major fonction of the system. Given that the local disks and their filesystems are accessible, that leaves us with: 1) Network. Not having one in working condition is a major setback these days. Network breakage can be avoided by appending needed modules to the /etc/initramfs-tools/modules, assuming one uses MODULES=dep setting in the /etc/initramfs-tools/initramfs.conf. And if one uses MODULES=most it's already done anyway (and MODULES=most is the default). 2) Video, sound, USB, etc. Painful, but bearable. There's nothing in those modules that can not be fixed by using snapshot.debian.org while having mounted root filesystem and a network. Still I agree that copying kernel and initrd does not substitute the backup. Reco -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: https://lists.debian.org/20140808061254.GA18207@x101h
